Thore Graepel

I have spent a working life teaching machines to reason under uncertainty and to act in the world. I trained as a physicist — in Hamburg, at Imperial College London, and in Berlin, where I earned my PhD in machine learning in 2001. At Microsoft Research from 2003 I helped found the Online Services and Advertising group, where reasoning under uncertainty shipped at scale: TrueSkill, the rating system behind matchmaking on Xbox Live, and AdPredictor, the click-through model behind Bing. Both rest on message passing over factor graphs.

At DeepMind I came back to my first love — building thinking machines — and had the luck to work on AlphaGo, the first program to beat a professional at the full-sized game of Go, and on its heirs AlphaGo Zero, AlphaZero and MuZero. There, learned reasoning met deep search: a policy that suggests, a value that weighs, a tree that looks ahead.

Much of my work since has dealt with multi-agent systems — from Capture-the-Flag agents to simulated humanoid football — where intelligence takes on a body, works alongside others, and grows through self-play. After leading machine learning for cellular rejuvenation at Altos Labs and a spell on Google DeepMind's Post-AGI team, those threads now meet in a new project: bringing AlphaGo-style reasoning to frontier AI, so that machines can plan and act under uncertainty. Search and learned judgement care little where the uncertainty springs from — a game tree, a cell, an agent choosing what to do next — so the same reasoning reaches on to embodied intelligence, and the multi-agent work above gives me grounds to think it will hold. Alongside this I hold the Chair of Machine Learning at UCL.